European AI Data Centers Hit Energy and Cooling Bottlenecks

Rapid growth of AI workloads is pushing data‑center designs in Europe to new physical limits. Modern AI clusters now exceed 100 kW per rack, and projections suggest individual racks could reach up to 1.2 MW by 2028. The surge in power density strains electricity supply, cooling systems and grid connections, turning energy efficiency into a strategic competitive factor for operators.

The EU AI Act is calling for greater transparency on AI‑related energy use, while researchers at the Fraunhofer Institute for Integrated Systems (Fraunhofer IIS) are testing quantum‑computing‑based training methods in the QC‑Train project. By moving the most resource‑intensive calculations onto quantum processors, the project aims to lower the electricity and cooling demand of large language‑model training, offering a potential pathway to more sustainable AI development.